We present observations of rovibrational CO in HD 100546 from four epochs spanning January 2003 through December 2010. We show that the equivalent widths of the CO lines vary during this time period with the v=1-0 CO lines brightening more than the UV fluoresced lines from the higher vibrational states. While the spectroastrometric signal of the hot band lines remains constant during this period, the spectroastrometric signal of the v=1--0 lines varies substantially. At all epochs, the spectroastrometric signals of the UV fluoresced lines are consistent with the signal one would expect from gas in an axisymmetric disk. In 2003, the spectroastrometric signal of the v=1-0 P26 line was symmetric and consistent with emission from an axisymmetric disk. However, in 2006, there was no spatial offset of the signal detected on the red side of the profile, and in 2010, the spectroastrometric offset was yet more strongly reduced toward zero velocity. A model is presented that can explain the evolution of the equivalent width of the v=1-0 P26 line and its spectroastrometric signal by adding to the system a compact source of CO emission that orbits the star near the inner edge of the disk. We hypothesize that such emission may arise from a circumplanetary disk orbiting a gas giant planet near the inner edge of the circumstellar disk. We discuss how this idea can be tested observationally and be distinguished from an alternative interpretation of random fluctuations in the disk emission.
Download